Integrating machine learning and data envelopment analysis for reliable reservoir water quality index assessment considering uncertainty

Water quality assessment is crucial for environmental health and quality of life. This study introduces a novel water quality index (WQI) model for reservoirs, using Wadi Dayqah Dam in Oman as a case study. The model advances water quality assessment using a data-driven approach, reducing reliance on subjective expert opinions. A large dataset of water samples was analysed using machine learning (ML) to select water quality variables (WQVs). Using a bootstrapping and subsampling approach, the proposed WQI was then calculated through sub-indexing, weighting, and aggregating sub-indices. WQV weights were estimated using gradient boosting and rank order centroid techniques, while aggregation involved scoring and data envelopment analysis (DEA). The model effectively captures uncertainty, prioritizes WQVs, and provides solutions to issues such as eclipsing, ranking, and dealing with bad variable values. The results were validated through uncertainty and sensitivity analyses, highlighting the model's potential for enhancing data-driven decision making in reservoir management.